Exporting a custom image to Cloud Storage

Linux VMs Windows VMs

If you need to move your Compute Engine boot disk data outside of your Compute Engine project, you can export a boot disk image to Cloud Storage as a tar.gz file. If you need to create a persistent disk image to use when you create new persistent disks on Compute Engine, read Creating a custom image.

You can backup or share a custom image by exporting the image to Cloud Storage. This method is ideal for sharing individual images with projects that do not have access to your images. Alternatively, you can share images by granting the Compute Engine image user role on the image or on the project that contains it.

The following diagram shows some typical workflows for the creation and reuse of a custom image.

Creating and reusing custom images. — Figure 1. Examples of creation and reuse of custom images

Before you begin

If you want to use the command-line examples in this guide:
1. Install or update to the latest version of the gcloud command-line tool.
2. Set a default region and zone.
If you want to use the API examples in this guide, set up API access.
Read the images page.
If the project that you want to export the image from has a trusted image policy defined, add projects/compute-image-tools to the allowed list of publishers.

Enable the Cloud Build API

The images export tool uses Cloud Build.

In most cases, gcloud compute images export attempts to grant these permissions to the Cloud Build service account. However, you can manually grant these permissions to ensure that the required permissions are in effect.

Console

In the Google Cloud Console, enable the Cloud Build API.
Enable the Cloud Build API

When you enable the Cloud Build API from the console, Compute Engine grants the Cloud Build service account the following roles so that the Cloud Build service can export instances from Compute Engine:
- roles/iam.serviceAccountTokenCreator
- roles/compute.admin
- roles/iam.serviceAccountUser
The export tool also uses the default Compute Engine service account. By default, the Compute Engine service account has the IAM project editor role. If this role is removed, the export process might fail. To add the role back to the service account, see Granting access. For more information about the Compute Engine default service account, see Compute Engine default service account.

gcloud

To set up the Cloud Build service using gcloud command-line tool, complete the following steps:

Using the gcloud command-line tool, enable Cloud Build.
```
gcloud services enable cloudbuild.googleapis.com
```
The export tool also uses the default Compute Engine service account. By default, the Compute Engine service account has the IAM project editor role. If this role is removed, the export process might fail. To add the role back to the service account, see Granting access. For more information about the Compute Engine default service account, see Compute Engine default service account.

Add the compute.admin role to the service account for the Cloud Build API.

gcloud projects add-iam-policy-binding PROJECT_ID \
   --member serviceAccount:PROJECT_NUM@cloudbuild.gserviceaccount.com \
   --role roles/compute.admin

Add the iam.serviceAccountUser role to the service account for the Cloud Build API.

gcloud projects add-iam-policy-binding PROJECT_ID \
   --member serviceAccount:PROJECT_NUM@cloudbuild.gserviceaccount.com \
   --role roles/iam.serviceAccountUser

Add the iam.serviceAccountTokenCreator role to the service account for the Cloud Build API.
```
gcloud projects add-iam-policy-binding PROJECT_ID \
   --member serviceAccount:PROJECT_NUM@cloudbuild.gserviceaccount.com \
   --role roles/iam.serviceAccountTokenCreator
```
Replace the following:
- PROJECT_ID: the projectID for your project.
- PROJECT_NUM: the project number for your project.

Limitations and restrictions

For projects that are protected with VPC Service Controls, only the manual method is supported.

Exporting an image with a single command

You can export your virtual disks using either the Google Cloud Console, the gcloud command-line tool, or the Cloud Build API.

Console

In the Google Cloud Console, go to the Images page.

Go to Images
Click the name of the image that you want to export to go to the image details page. You can't export public images provided by Google. You can only export images that you previously created or imported.
From the image details page, click Export to open the Export Image page.
From the Export image page, choose the Export format of the image.
Choose the Cloud Storage location to export your image to by clicking Browse.
Choose an existing Cloud Storage location to export your image. Or, follow the directions to create a new Cloud Storage bucket, and then enter a name for the new Cloud Storage bucket.
Once you choose a Cloud Storage, choose a filename for the exported image. You can use the default filename, or you can choose your own filename.
After choosing a Cloud Storage, and entering a filename for the image, click Select.
From the Export image page, click Export. After choosing Export, the Cloud Console displays the Image export history, where you can view the image export process. For additional details about the image export process, click the Cloud Build ID to go to the Image export details page where you can view and download the image export log.
Go to the Storage page to access your exported image.

Go to Storage

gcloud

The preferred way to export an image to Cloud Storage is to use the gcloud compute images export command. This command uses Daisy to chain together the multiple steps that are required to export an image. It assumes that you have already created an image, for example, with the gcloud compute images create command.

Using the gcloud command-line tool, run:

gcloud compute images export \
    --destination-uri DESTINATION_URI \
    --image IMAGE_NAME

Replace the following:

DESTINATION_URI: the Cloud Storage URI destination for the exported virtual disk file.
IMAGE_NAME: the name of the disk image to export.

Example

For example, the following command exports an image named my-image from my-project to a Cloud Storage bucket named my-bucket. By default, the image is exported as a disk.raw file and is compressed into the tar.gz file format.

gcloud compute images export \
    --destination-uri gs://my-bucket/my-image.tar.gz \
    --image my-image \
    --project my-project

For flags, see the gcloud compute images export reference documentation.

API

In the API, create a POST request to the Cloud Build API.

Note: Two timeout values are required. Default values for these are provided in the following sample requests. These default values work in most scenarios. However, if you need to adjust these values, ensure that the Cloud Build timeout is at least 3% greater than the timeout specified for the arguments (args) section.

POST https://cloudbuild.googleapis.com/v1/projects/PROJECT_ID/builds
{
  "timeout": "7200s",
  "steps":[
    {
      "args":[
        "-timeout=7000s",
        "-source_image=SOURCE_IMAGE",
        "-client_id=api",
        "-format=IMAGE_FORMAT",
        "-destination_uri=DESTINATION_URI"
      ],
      "name":"gcr.io/compute-image-tools/gce_vm_image_export:release",
      "env":[
        "BUILD_ID=$BUILD_ID"
      ]
    }
  ],
  "tags":[
    "gce-daisy",
    "gce-daisy-image-export"
  ]
}

Replace the following:

PROJECT_ID: the project ID for the project that contains the image that you want to export.
SOURCE_IMAGE: the name of the image to be exported.
IMAGE_FORMAT: the format of the exported image. Valid formats include vmdk, vhdx, vpc, vdi, and qcow2.
DESTINATION_URI: the Cloud Storage URI location that you want to export the virtual disk file to. For example, gs://my-bucket/my-exported-image.vmdk.

For additional args values that can be provided, see the optional flags section of the VM image export GitHub page.

Example response

The following sample response resembles the output that is returned:

{
"name": "operations/build/myproject-12345/operation-1578608233418",
"metadata": {
 "@type": "type.googleapis.com/google.devtools.cloudbuild.v1.BuildOperationMetadata",
 "build": {
  "id": "3a2055bc-ccbd-4101-9434-d376b88b8940",
  "status": "QUEUED",
  "createTime": "2019-10-02T18:59:13.393492020Z",
  "steps": [
   {
    "name": "gcr.io/compute-image-tools/gce_vm_image_export:release",
    "env": [
     "BUILD_ID=3a2055bc-ccbd-4101-9434-d376b88b8940"
    ],
    "args": [
     "-timeout=7056s",
     "-source_image=my-image",
     "-client_id=api",
     "-format=vmdk",
     "-destination_uri=gs://my-bucket/my-exported-image.vmdk"
    ]
   }
  ],
  "timeout": "7200s",
  "projectId": "myproject-12345",
  "logsBucket": "gs://123456.cloudbuild-logs.googleusercontent.com",
  "options": {
   "logging": "LEGACY"
  },
  "logUrl": "https://console.cloud.google.com/gcr/builds/3a2055bc-ccbd-4101-9434-d376b88b8940?project=123456"
 }
 }

There are a couple ways you can monitor your build:

Run a projects.builds.get request using the returned build-id.
Review the logs hosted at the provided logUrl.

Exporting an image from a project using a custom Compute Engine service account

During an image export, a temporary virtual machine (VM) instance is created in your project. The image export tool on this temporary VM must be authenticated.

A service account is an identity that is attached to a VM. Service account access tokens can be accessed through the instance metadata server and used to authenticate the image export tool on the VM.

By default, the export process uses the project's default Compute Engine System service account. However, if the default Compute Engine service account is disabled in your project or if you want to use a custom Compute Engine service account, then you need to create a service account and specify it for the export process.

gcloud

Create a service account and assign the minimum roles. For more information about creating service accounts, see Creating and managing service accounts.

At minimum, the specified Compute Engine service account needs to have the following roles assigned:
- roles/compute.storageAdmin
- roles/storage.objectAdmin
Use the gcloud compute images export command to export the image.
```
gcloud compute images export \
    --destination-uri DESTINATION_URI \
    --image IMAGE_NAME \
    --compute-service-account SERVICE_ACCOUNT_EMAIL
```
Replace the following:
- DESTINATION_URI: the Cloud Storage URI destination for the exported virtual disk file.
- IMAGE_NAME: the name of the disk image to export.
- SERVICE_ACCOUNT_EMAIL: the email address associated with the Compute Engine service account created in the previous step.

Example

For example, the following command exports an image named my-image from my-project to a Cloud Storage bucket named my-bucket with a service account that has the email image-export-service-account@proj-12345.iam.gserviceaccount.com. By default, the image is exported as a disk.raw file and is compressed into the tar.gz file format.

gcloud compute images export \
    --destination-uri gs://my-bucket/my-image.tar.gz \
    --image my-image \
    --project my-project \
    --compute-service-account image-export-service-account@proj-12345.iam.gserviceaccount.com

For flags, see the gcloud compute images export reference documentation.

API

Create a service account and assign the minimum roles. For more information about creating service accounts, see Creating and managing service accounts.

At minimum, the specified Compute Engine service account needs to have the following roles assigned:
- roles/compute.storageAdmin
- roles/storage.objectAdmin
In the API, create a POST request to the Cloud Build API.

Note: Two timeout values are required. Default values for these are provided in the following sample requests. These default values work in most scenarios. However, if you need to adjust these values, ensure that the Cloud Build timeout is at least 3% greater than the timeout specified for the arguments (args) section.
```
POST https://cloudbuild.googleapis.com/v1/projects/PROJECT_ID/builds
{
  "timeout": "7200s",
  "steps":[
    {
      "args":[
        "-timeout=7000s",
        "-source_image=SOURCE_IMAGE",
        "-client_id=api",
        "-format=IMAGE_FORMAT",
        "-destination_uri=DESTINATION_URI",
        "-compute_service_account=SERVICE_ACCOUNT_EMAIL"
      ],
      "name":"gcr.io/compute-image-tools/gce_vm_image_export:release",
      "env":[
        "BUILD_ID=$BUILD_ID"
      ]
    }
  ],
  "tags":[
    "gce-daisy",
    "gce-daisy-image-export"
  ]
}
```
Replace the following:
- PROJECT_ID: the project ID for the project that contains the image that you want to export.
- SOURCE_IMAGE: the name of the image to be exported.
- IMAGE_FORMAT: the f ormat of the exported image. Valid formats include vmdk, vhdx, vpc, vdi, and qcow2.
- DESTINATION_URI: the Cloud Storage URI location that you want to export the virtual disk file to. For example, gs://my-bucket/my-exported-image.vmdk.
- SERVICE_ACCOUNT_EMAIL: the email address associated with the Compute Engine service account created in the previous step.

For additional args values that can be provided, see the optional flags section of the VM image export GitHub page.

Creating and exporting an image manually

If the gcloud compute images create and gcloud compute images export commands do not meet your requirements, you can create and export an image manually from a Compute Engine instance. This process has discrete steps to first create an image and then export an image.

In the following example, note the created disk is called image-disk.

To create and export an image:

Optionally, stop the instance that the disk is attached to before you create the snapshot. Stopping the instance ensures the integrity of the disk contents in the snapshot. Replace DISK_NAME with the name of the disk that you want to use to create the snapshot.

Create a snapshot of the disk. Name the snapshot image-snapshot.

gcloud compute disks snapshot DISK_NAME \
    --snapshot-names image-snapshot

Use the image-snapshot snapshot to create a new disk named image-disk by running the following command:
```
gcloud compute disks create image-disk \
    --source-snapshot image-snapshot
```
Create a temporary disk named temporary-disk to hold your tar file, and specify the SIZE of the disk to be at least 50% larger than the image disk.

You can detach and delete the disk afterwards.
```
gcloud compute disks create temporary-disk \
    --size SIZE
```
where SIZE is the size, in gigabytes or terabytes, of the temporary disk. For example, specify 100GB to create a 100-gigabyte disk.
Create an instance and enable storage-rw scope on the instance. Also, attach the image-disk and the temporary-disk to the instance as secondary disks with specific device-name attributes. Replace VM_NAME with the name of the instance to create.
```
gcloud compute instances create VM_NAME \
    --scopes storage-rw \
    --disk name=image-disk,device-name=image-disk \
    --disk name=temporary-disk,device-name=temporary-disk
```
Note that you're passing in service account scopes so that you can upload your file to Cloud Storage in later steps.

Review the details about starting a new instance if necessary.
Connect to your instance. Replace VM_NAME with the name of the instance to connect to.
```
gcloud compute ssh VM_NAME
```

Format and mount the temporary disk. Formatting the disk deletes the contents of the temporary disk.

sudo mkdir /mnt/tmp

sudo mkfs.ext4 -F /dev/disk/by-id/google-temporary-disk

sudo mount -o discard,defaults /dev/disk/by-id/google-temporary-disk /mnt/tmp

Optionally, you can mount the image disk and make additional changes before you create the tar file. For example, you might want to delete any existing files from the /home directory if you do not want them to be part of your image. Mount the disk partitions that you need to modify, modify the files on the disk that you need to change, and then unmount the disk when you are done.
1. Create a directory where you can mount your disk or partition.
```
sudo mkdir /mnt/image-disk
```
2. Use the ls command to determine which disk or disk partition you need to mount.
```
ls /dev/disk/by-id/
```
  The command prints a list of disk IDs and partitions. For example, the following disk has a partition table with one partition. The google-image-disk ID points to the full disk from which you want to create an image. The google-image-disk-part1 ID points to the first partition on this disk. Mount the partition if you need to make changes to the disk, then create the image from the full disk.
```
google-image-disk
google-image-disk-part1
```
3. Mount the disk or the partition. If your disk has a partition table, mount the individual partitions for your disk. For example, mount google-image-disk-part1.
```
sudo mount /dev/disk/by-id/google-image-disk-part1 /mnt/image-disk
```
  Alternatively, if your disk is raw formatted with no partition table, mount the full google-image-disk disk.
```
sudo mount /dev/disk/by-id/google-image-disk /mnt/image-disk
```
4. Modify the files in the /mnt/image-disk directory to configure the files on the disk. As an example, you might remove the /mnt/image-disk/home/[USER]/.ssh/authorized_keys file to protect your SSH keys from being shared.
5. After you have finished modifying files, unmount the disk.
```
sudo umount /mnt/image-disk/
```
Create a tar file of your image.

When you finish customizing the files on the image disk, create a raw disk file on your temporary disk. The name of the raw disk image must be 'disk.raw':
```
 sudo dd if=/dev/disk/by-id/google-image-disk of=/mnt/tmp/disk.raw bs=4096
```
Then tar and gzip this file:
```
cd /mnt/tmp
```
```
sudo tar czvf myimage.tar.gz disk.raw
```
This command creates an image of the instance in the following location:
```
/mnt/tmp/myimage.tar.gz
```
Upload the image into Cloud Storage.

To upload the tar file to Cloud Storage, use the gsutil command line tool that comes preinstalled on your instance.
1. Create a bucket using gsutil.
  
  Make sure to review the bucket and object naming guidelines before you create your bucket. Then, create your bucket using the following command. Replace BUCKET_NAME with the name of the bucket to create.
```
me@example-instance:~$ 
gsutil mb gs://BUCKET_NAME
```
2. Copy your file to your new bucket. Replace BUCKET_NAME with the name of the bucket to copy the file to.
```
me@example-instance:~$ 
gsutil cp /mnt/tmp/myimage.tar.gz gs://BUCKET_NAME
```

You have exported your file into Cloud Storage. You can now share the image with other people, or use the tar file to add a new image to a Google Cloud Console project.

What's next

Share images using the image user role.
Learn about the import methods available for Compute Engine.